A voltage-source converter is connected to a direct-voltage network which occurs as a stiff voltage source towards the converter. This property of the direct-voltage network is often achieved by connecting a capacitor between the poles of the network. It is known to design such converters comprising gate turn-off semiconductor devices, usually connected in a bridge connection. In these cases, the direct voltage of the converter may in some way known per se be controlled by means of pulse-width modulation. The technique comprising pulse-width modulation makes it possible to eliminate the transformer which is normally arranged between the alternating-voltage terminals of the converter and the alternating-voltage network, which entails a considerable saving of cost. However, because of its mode of operation, a bridge of the above-mentioned kind only permits two-pole operation, by which is meant operation with the two direct-voltage terminals of the converter connected to poles on a direct-voltage network which are located at an essentially symmetrical direct voltage in relation to a neutral potential, usually ground potential. This is a serious limitation since, for example in installations for high-voltage direct current with converter stations designed for two-pole operation, in the event of a fault on one of the poles, it is a requirement to be able to maintain single-pole operation, by which is meant operation with one pole in the direct-voltage network at a potential substantially corresponding to the potential of the neutral pole. In installations designed for single-pole operation only, the requirement for a connection of the converter to the alternating-voltage network via a transformer, which requirement is due to technical reasons, is a considerable disadvantage from the point of view of cost.